dontbesilent2025/dbskill

dbskill is a set of 17 AI agent skills for business diagnosis, built by a Chinese creator who distilled methods from 12,307 of their own tweets. The skills are designed to be loaded into AI coding tools like Claude Code, Codex, Cursor, or Trae Solo, and each one guides the AI through a specific type of analysis rather than answering questions directly. The core diagnostic tools walk through distinct stages of figuring out what is wrong with a business direction, content strategy, or personal goal. One skill diagnoses business model problems, another finds reference points by comparing against similar successful operators, a third evaluates content across five dimensions, and others cover specific concerns like short-video openings, Xiaohongshu (a Chinese social platform) titles, detecting AI-sounding writing, and whether someone is taking shortcuts on a decision that deserves slow careful work. A state management trio rounds out the toolkit. After finishing a diagnosis session, the user can save the key conclusions to a local file, reload them in a later session to continue where they left off, and eventually merge multiple saved sessions into a single shareable report. This is meant to turn diagnosis from a one-time chat into an ongoing process across days or weeks. Alongside the skills is an open knowledge base. It contains 4,176 structured knowledge atoms, each one a single insight extracted from the author's tweets and formatted as JSON with topic tags, a confidence level, and links to the original post. The same knowledge base also includes longer methodology documents and case libraries for each skill area. These files can be used independently, for example as a system prompt for a custom chatbot or as a dataset for retrieval-augmented search, without installing the skill package at all. The project is licensed under CC BY-NC 4.0, which allows personal, educational, and non-commercial use freely. Commercial use requires separate permission from the author.